基于在线学习的自适应粒子滤波视频目标跟踪

传统的粒子滤波视觉跟踪算法采用固定模型和大量粒子表征目标后验概率,不能满足复杂条件下的视频目标实时跟踪.为了提高跟踪的鲁棒性和稳定性及计算效率,本文提出将自适应状态演化方程和在线增量学习观测似然模型嵌入到粒子滤波算法;并采用在线自动调整粒子数目的策略,提高粒子滤波视觉跟踪的计算效率.室内外实验结果表明,文中提出的视觉跟踪算法不仅能准确、高效地跟踪序列图像中的运动目标,而且对光照、姿态变化引起的目标表观变化具有良好的鲁棒性.     

基于高斯粒子滤波的当前统计模型跟踪算法

对于非线性系统估计问题,高斯粒子滤波器可以获得近似最优解,与粒子滤波器相比其优点是不需要重采样步骤和不存在粒子退化现象.采用高斯粒子滤波代替当前模型自适应跟踪算法中的卡尔曼滤波,将高斯粒子滤波与当前统计模型的优点相结合,提出了一种新的当前统计模型自适应跟踪算法,用于非线性非高斯系统的机动目标跟踪.MonteCarlo仿真表明,该算法跟踪精度优于标准的交互多模型算法和当前统计模型自适应跟踪算法,实时性好于交互多模型粒子滤波算法.     

基于改进的粒子群优化的FastSLAM方法

提出了一种基于改进的粒子群优化(IPSO)的快速同时定位和地图创建(FastSLAM)方法--IPSO FastSLAM算法.该算法在粒子预估过程中引入观测信息,调整了粒子的提议分布,增强了位置预测的准确性.改进的粒子群优化采用两步优化策略,即首先通过种群速度自适应调整惯性权重,有效地克服了粒子退化问题,改善了算法的实...     

基于关节层次模型的3D手臂运动快速跟踪算法

针对人体3D运动跟踪过程中计算效率低下的问题,提出了基于粒子滤波和关节层次模型的3D手臂运动快速跟踪算法.该算法在经典的人体3D骨骼模型的基础上,提出了关节层次模型,利用关节层次模型将高维手臂运动状态参数空间分解为若干个低维参数子空间,最后,在关节层次模型的约束条件下,根据自顶向下的搜索策略,利用粒子滤波算法对手臂运动进行跟踪,从而减少了跟踪所需的粒子数目.实验表明,与标准的粒子滤波算法相比,该算法能够在保证跟踪精度的同时,提高计算效率.     

基于水声模型多层次并行计算的匹配反演典型应用研究

海洋环境参数的时变特性给海洋环境参数进行直接测量带来了很多挑战，目前主流的方法是通过反演获得海洋环境参数。粒子滤波是一种重要的匹配反演方法。该方法基于贝叶斯定理，通过对海洋环境参数充分采样，能够较精确地计算出环境参数的后验概率，从而取得较好的反演性能。然而粒子滤波匹配反演方法随着粒子数的增大，计算量也急剧增大，为此，文章提出了基于水声模型的多层次并行方法，能够将粒子滤波算法高效映射到多核集群的硬件体系结构中。最后在天河2号超级计算平台进行了粒子滤波算法的并行性能测试，在单节点多核并行测试中取得了87.5%的并行效率，在多节点强扩展测试中，粒子数达到12 288个，在128个计算节点中取得了近110倍的加速性能。     

基于遗传进化策略的粒子滤波视频目标跟踪

粒子退化问题是影响基于粒子滤波视觉跟踪性能的一个重要因素,为克服这种缺陷,本文将遗传进化策略引入粒子滤波跟踪算法,利用遗传算法的选择策略,根据预定的似然阈值迭代选择每代粒子中次优个体,然后对未选中的粒子实施交叉、变异操作以获得粒子的多样性,从而有效解决了粒子的退化问题.另外,针对跟踪中目标表观变化的问题,本文提出的跟踪算法采用了多模板自适应更新技术以确保跟踪的准确性.实验结果表明,本文提出的跟踪算法能有效地跟踪室内运动目标,并对光照变化、目标姿态变化具有良好的鲁棒性.     

结合颜色和结构信息的粒子滤波跟踪算法

单纯依靠颜色信息往往使得目标跟踪不稳定.为了在复杂背景和光照变化等干扰情况下,能够准确地跟踪到感兴趣的目标,本文提出了将颜色和结构信息相结合的方法.该方法在基于颜色直方图的粒子滤波跟踪框架中,利用目标的灰度图像建立结构模型,通过结构相似性质量因子衡量目标在两帧图像之间的相似性.实验表明,该算法在使用相同粒子数目的情况下较传统的基于颜色直方图的粒子滤波跟踪算法鲁棒性更高,可以用于特定场合可疑目标的跟踪.     

改进型滤波算法在金刚石检测处理中的应用

提出一种改进型自适应滤波算法,该算法结合图像灰度直方图,先将噪声限制在一定的灰度值范围内,然后通过调整窗口局部大小进行滤波处理。仿真结果表明,与其他滤波算法相比,该方法的处理效果明显改善。     

基于图像去噪的混合噪声滤波算法

传统的中值滤波和均值滤波通常被分别用来滤除椒盐噪声和高斯噪声.但是当图像同时存在高斯噪声和椒盐噪声时,单独使用哪种滤波方法都不会达到最好的去噪效果.为了能同时滤除两种不同性质的噪声,提出了一种新的自适应混合噪声滤波算法.该算法采用了一种基于自适应阈值的方法对滤波系数加以优化,使其既可以有效地减少噪声,又可以较好的保持图像的边缘细节信息,仿真结果表明该算法能较好的滤除混合噪声,且滤波效果优于传统的滤波算法.     

一种改进的自适应重采样粒子滤波算法

针对粒子滤波跟踪算法中重采样所引起的粒子多样性缺失问题,提出了一种自适应粒子滤波重采样方法。首先将粒子权值进行分类,中等权值粒子保持不变,大、小权值粒子进行权值优化组合,其次对优化组合后的小部分粒子进行系统重采样。最后对粒子的权值及粒子的复制次数分别进行比较运算。试验结果表明,改进后的粒子滤波算法不仅提高了机动目标跟踪的运算效率,而且还有效提高了跟踪的稳定性,目标跟踪更加准确。     

Estimation of particle size distribution from cross-sectional particle diameter on the cutting plane

To control highly functional sintered materials, it is necessary to evaluate particle size segregation within materials. In the present study, a new method for estimating particle size distribution is proposed; this method considers the occurrence probability of the cutting diameter. The proper particle size distribution in a particle bed was estimated by calculating a matrix consisting of the occurrence probability and the distribution of particle diameters measured on a cutting plane. The estimated particle size distribution was smoothed using the Phillips–Twomey method. A cavity-filling simulation was carried out to verify the validity of the proposed method using the Distinct Element Method. The particle size distribution estimated by this method correlated well with the actual particle size distribution. The effect of particle size distributions with various geometrical standard deviations on the accuracy of estimated values was also investigated. The accuracy increased as the geometric standard deviation increased, and there was an optimum particle size bin number for a specific particle distribution. It was found that a large bin number and a large number of measured particles were required to obtain a higher accuracy for narrow size distributions.     

Alternative approach for defining the particle population requirements for static image analysis based particle characterization methods

For image based particle characterisation approaches one of the most common discussion points is determining the number of particles required to have statistical confidence that the measurement is able to adequately describe the distribution of the sample. This topic becomes significantly more challenging when applied to the extraction of single component size distributions from multi-component samples.The aim of this work was to propose a means to accurately assess the particle number requirements using a method specific approach. The method applies a sub-sampling method to the original imaged dataset in order to provide an understanding of the impact of sub-sampling on the ability to accurately reproduce the original distribution.The method was applied to understand the particle number requirements for two batches of theophylline anhydrous with varied particle size distributions, using the input size distribution to guide the requirements for the subsequent multi-component samples of both materials.The results demonstrate the utility of the method to determine the appropriate number of particles required to recreate the size distributions. Whilst the minimum number of particles required to be sampled can be estimated, how those particles are sampled can also affect the validity of the measurement and must be taken into consideration.     

氧化铝和氢氧化铝颗粒分析方法

简述了激光衍射法测量氧化铝和氢氧化铝颗粒粒度的原理和方法，以及其分析方法的优缺点，通过在砂状氧化铝试验过程中对大量的氧化铝和氢氧化铝颗粒粒度测量和分析，总结了影响颗粒粒度分析的共性因素。提出消除影响因素的措施和改进氧化铝和氢氧化铝颗粒粒度的分析方法。     

The development of the measurement of particle concentration using a commercial laser diffraction particle size analyzer

A laser diffraction particle size analyzer based on the Fraunhofer diffraction theory has the advantages of real-time measurement of particle size distribution over a broad range. However, the dispersed particle number concentration is not displayed in commercially available analyzers. The method of measuring the dispersed particle number concentration was investigated for different particles having various shapes, i.e. spherical, cubical and prismoidal with a log-normal distribution, by considering the characteristics of the measured voltage of the set detector in relation to the diffracted light intensity using a commercial laser diffraction particle size analyzer. As a result, an approximate equation for the particle number concentration was proposed expressing the measured median diameter and the highest voltage measured by the detector. This equation was applicable to particles having various shapes, i.e. spherical, cubical and prismoidal. Furthermore, this technique can be used for the continuous measurement of the particle number concentration of growing crystals in the crystallizer for crystallization operation.     

微观图像技术在超细材料粒度分析中的应用

应用计算机微观图像采集分析系统建立了一套对超细材料进行形状表征和粒度测量的方法。该方法克服以往图像法测量样本小 ,得不到正确粒度分布的弊端。通过对SiO2 标准物的粒度测定表明 :均匀、单分散的测量样片的制备是图像法准确测量和分析的关键。当测量图片在 6张以上 ,累积测量颗粒总数在 2 0 0个左右时 ,测量结果具有代表性。该系统用于测量的粒度范围为 0 .5 μm～ 1cm的粒子时 ,长度测量误差小于 5 % ,颗粒等面积直径测量偏差小于12 %。     

An application of bootstrap method for analysis of particle size distribution

The Bootstrap method, which is used widely in statistics, is a very powerful method that can be applied to the analysis of particle size distribution. For a number based measurement, this method can estimate the statistic uncertainty or confidence interval for any statistical quantities of interest from the distribution, with very simple protocol and without any parametric assumptions. This paper gives a demonstration to introduce the method to the community of particle size analysis.     

Quantitative simulation of gas-particle two phase plane mixing layer using discrete vortex method

The gas and particle phase in a two-phase plane mixing layer flow are numerically simulated using the discrete vortex method and a trajectory tracking method. It is shown that the number of vortex elements contained in two semi-infinite discrete vortex sheet and the method of generating control volumes for statistical calculation of the particle phase have important effects on the predicted results of particle phase, especially for quantitative prediction. By adopting different number of vortex elements for two semi-infinite discrete vortex sheet and overlapping the control volumes, predicted results including streamwise velocity, fluctuating velocity and Reynolds shear stress of both phases are obtained and agree well with experimental measurements quantitatively. It shows that the discrete vortex method can achieve the accurate quantitative simulation of two-phase flow.     

Mass-measurement-type optical particle counting method for determination of number concentration of liquid-borne particles

We have developed a measurement method that enables accurate determination of the number concentration of liquid-borne particles by adopting simultaneous particle counting and mass measurement. The new measurement method consists of an optical particle counter (OPC) for particle counting, an electronic balance for mass measurement, a syringe pump for sampling, and a pulse height analyzer for signal processing. We refer to this method as the mass-measurement-type optical particle counting (M-OPC) method. The M-OPC method facilitates both metrological traceability of the mass of measured suspensions and evaluation of the counting efficiency of OPCs. Using the M-OPC method, we evaluated the measurement accuracy of two different types of OPCs: a light-scattering type and a light-extinction type. The validity of the measurement results obtained with the M-OPC method was examined by comparison with the results of an independent microscopic method using a scanning electron microscope (SEM). The M-OPC method also offers the possibility of calibrating the number concentration of particles in suspensions with diameters from the sub-micrometer to micrometer range by combining multiple OPCs. This method is expected to be used as the national primary standard at the National Metrology Institute of Japan of the National Institute of Advanced Industrial Science and Technology.     

A non-obstructive particle damping model of DEM

The non-obstructive particle damping technology has been used successfully in many fields for vibration reduction. However, it is difficult to predict the damping characteristics due to complex collisions in the dense particle flow. The discrete element method makes it possible to consider effects of granularity such as the particle size, number of particles and friction between two particles. The validity of this numerical method is examined by a comparison of the experimental results. The discrete element method simulation system is further examined with examples for its computational complexity and effectiveness for different density parameters. It is shown that the mass ratio and material density influence the damping performance.